Optimization of Energy Consumption in Electric Vehicles based on Wasserstein Deep Conditional Generative Adversarial Network with Particle Swarm Optimization

From the past few years, Battery Management Systems (BMS) in electric vehicles plays a significant role in transportation industry. However, existing Siberian Tiger Optimization and an enhanced Wasserstein Generative Adversarial Network (STO-IWGAN) model has poor generalization performance on unseen data, leading to overfitting and degraded performance. To overcome this challenge, this research proposes a Wasserstein Deep Conditional GAN with Particle Swarm Optimization (WDCGAN-PSO) for enhancing energy optimization in electric vehicles and reduce the risk of overfitting. Initially, data is collected from Advanced Driver Assistance Systems in Electric Vehicles (ADAS-EV) dataset, which is comprehensive collection of real-time sensor data. Further, it preprocessed with Butterworth Filters technique which multiples the input signal with filter transfer function to remove high-frequency noise and reduces interference between signals. After that, Short-Time Fourier Transform (STFT) is utilized to extract time-frequency features by decomposing short window signal through Discrete Fourier Transform (DFT) and fed into WDCGAN. This model generates the energy usage patterns to minimize energy consumption, then PSO technique optimizes by evaluating fitness of each particle using fitness function. From the results, the proposed WDCGAN-PSO model achieved outstanding results when compared with STO-IWGAN model in terms of accuracy (0.99), precision (0.98), and recall (0.97).